mlr3oml enables downloading OpenML datasets, tasks, flows, runs,
and collections and allows to convert them to the corresponding mlr3 object. Furthermore, you can
also obtain the data and the resampling for a given OpenML task. Caching
can be enabled by setting the option "mlr3oml.cache".
Uploading to OpenML is currently not supported, use the OpenML package
package for this.
library("mlr3")
library("mlr3oml")
# be less verbose
lgr::get_logger("mlr3oml")$set_threshold("warn")
# retrieve data set as task from OML
tsk("oml", data_id = 31)## <TaskClassif:credit-g> (1000 x 21)
## * Target: class
## * Properties: twoclass
## * Features (20):
## - fct (13): checking_status, credit_history, employment,
## foreign_worker, housing, job, other_parties, other_payment_plans,
## own_telephone, personal_status, property_magnitude, purpose,
## savings_status
## - int (7): age, credit_amount, duration, existing_credits,
## installment_commitment, num_dependents, residence_since# retrieve a regular task from OML
tsk("oml", task_id = 59)## <TaskClassif:iris> (150 x 5)
## * Target: class
## * Properties: multiclass
## * Features (4):
## - dbl (4): petallength, petalwidth, sepallength, sepalwidth# retrieve resampling from OML
rsmp("oml", task_id = 59)## <ResamplingCustom>: Custom Splits
## * Iterations: 10
## * Instantiated: TRUE
## * Parameters: list()# R6 class for data sets
odata = OMLData$new(61) # or
odata = odt(61)
odata$name## [1] "iris"odata$nrow## [1] 150odata$ncol## [1] 5head(odata$data)## sepallength sepalwidth petallength petalwidth class
## 1: 5.1 3.5 1.4 0.2 Iris-setosa
## 2: 4.9 3.0 1.4 0.2 Iris-setosa
## 3: 4.7 3.2 1.3 0.2 Iris-setosa
## 4: 4.6 3.1 1.5 0.2 Iris-setosa
## 5: 5.0 3.6 1.4 0.2 Iris-setosa
## 6: 5.4 3.9 1.7 0.4 Iris-setosabackend = as_data_backend(odata)
backend## <DataBackendDataTable> (150x6)
## sepallength sepalwidth petallength petalwidth class ..row_id
## 5.1 3.5 1.4 0.2 Iris-setosa 1
## 4.9 3.0 1.4 0.2 Iris-setosa 2
## 4.7 3.2 1.3 0.2 Iris-setosa 3
## 4.6 3.1 1.5 0.2 Iris-setosa 4
## 5.0 3.6 1.4 0.2 Iris-setosa 5
## 5.4 3.9 1.7 0.4 Iris-setosa 6
## [...] (144 rows omitted)# list oml data sets with 5 features and 50 - 200 instances
tab = list_oml_data(number_features = 5, number_instances = c(50, 200))
head(tab[, .(data_id, name)])## data_id name
## 1: 61 iris
## 2: 199 fruitfly
## 3: 214 baskball
## 4: 329 hayes-roth
## 5: 346 aids
## 6: 668 witmer_census_1980# R6 class for tasks
otask = OMLTask$new(31) # or
otask = otsk(31)
task = as_task(otask)
task## <TaskClassif:credit-g> (1000 x 21)
## * Target: class
## * Properties: twoclass
## * Features (20):
## - fct (13): checking_status, credit_history, employment,
## foreign_worker, housing, job, other_parties, other_payment_plans,
## own_telephone, personal_status, property_magnitude, purpose,
## savings_status
## - int (7): age, credit_amount, duration, existing_credits,
## installment_commitment, num_dependents, residence_sinceresampling = as_resampling(otask)
resampling## <ResamplingCustom>: Custom Splits
## * Iterations: 10
## * Instantiated: TRUE
## * Parameters: list()otask$data## <OMLData:31:credit-g> (1000x21)
## * Default target: classotask$name## [1] "Task 31: credit-g (Supervised Classification)"otask$nrow## [1] 1000otask$ncol## [1] 21otask$task## NULLotask$resampling## NULL# list first 10 oml tasks
tab = list_oml_tasks(limit = 10)
tab[, .(task_id, data_id, name)]## task_id data_id name
## 1: 2 2 anneal
## 2: 3 3 kr-vs-kp
## 3: 4 4 labor
## 4: 5 5 arrhythmia
## 5: 6 6 letter
## 6: 7 7 audiology
## 7: 8 8 liver-disorders
## 8: 9 9 autos
## 9: 10 10 lymph
## 10: 11 11 balance-scale# R6 class for flows
oflow = OMLFlow$new(100) # or
oflow = oflw(100)
oflow$dependencies## [1] "Weka_3.7.5"oflow$parameter## name data_type default_value
## 1: A flag
## 2: B flag
## 3: C option 0.25
## 4: J flag
## 5: L flag
## 6: M option 2
## 7: N option
## 8: O flag
## 9: Q option
## 10: R flag
## 11: S flag
## 12: U flag# non-executable pseudo learner
learner = as_learner(oflow, "regr")
learner## <LearnerRegrOML100:oml.100>
## * Model: -
## * Parameters: list()
## * Packages: mlr3
## * Predict Types: [response]
## * Feature Types: -
## * Properties: -# R6Class for run
orun = OMLRun$new(538858) # o
orun = orn(538858)
orun$data## <OMLData:952:prnn_fglass> (214x10)
## * Default target: typeorun$task## <OMLTask:3815>
## * Type: Supervised Classification
## * Data: prnn_fglass (id: 952; dim: 214x10)
## * Target: type
## * Estimation: crossvalidation (id: 1; repeats: 1, folds: 10)orun$flow## <OMLFlow:3364>
## * Name: classif.boosting
## * Dependencies: mlr_2.8, adabag_4.1, rpart_4.1.10head(orun$prediction)## repeat. fold row_id prediction truth
## 1: 0 0 33 window_float_glass window_float_glass
## 2: 0 0 52 window_float_glass window_float_glass
## 3: 0 0 67 window_float_glass window_float_glass
## 4: 0 0 49 window_float_glass window_float_glass
## 5: 0 0 41 window_float_glass window_float_glass
## 6: 0 0 5 window_non-float_glass window_float_glass
## confidence.containers confidence.tableware confidence.vehicle_glass
## 1: 0 0 0
## 2: 0 0 0
## 3: 0 0 0
## 4: 0 0 0
## 5: 0 0 0
## 6: 0 0 0
## confidence.vehicle_headlamp_glass confidence.window_float_glass
## 1: 0 1
## 2: 0 1
## 3: 0 1
## 4: 0 1
## 5: 0 1
## 6: 0 0
## confidence.window_non.float_glass
## 1: 0
## 2: 0
## 3: 0
## 4: 0
## 5: 0
## 6: 1rr = as_resample_result(orun)
rr## <ResampleResult> of 10 iterations
## * Task: prnn_fglass
## * Learner: oml.3364
## * Warnings: 0 in 0 iterations
## * Errors: 0 in 0 iterations# R6 class for collection
ocol = OMLCollection$new(232) # or
ocol = ocl(232)
ocol$tasks## id task data task_type target nrow ncol
## 1: 3 <OMLTask[25]> kr-vs-kp Supervised Classification class 3196 37
## 2: 6 <OMLTask[25]> letter Supervised Classification class 20000 17
## missing numeric symbolic binary task_splits
## 1: 0 0 37 35 crossvalidation
## 2: 0 16 1 0 crossvalidation